Re: Preliminary HDF5 Dimension documents

Robert E. McGrath wrote:


John Caron <caron@xxxxxxxxxxxxxxxx> writes:


we need shared dimensions: almost all higher level Conventions rely on them. they dont always rely on them for just coordinate variables. Generalizarions of coordinate variables also need shared dimensions, but cannot use coordinate variables.


John,

Pardon my ignorance.

Could you give me some pointers to the "higher level Conventions"
that we should try to target.

(I think you can see where I'm coming from: if there is a requirement
here, we need to state it carefully and make sure we are addressing it.)

Conventions are a way of adding semantics to what is really a pretty low-level data model. Heres a non-exhaustive list:

 http://www.unidata.ucar.edu/packages/netcdf/conventions.html

I dont know that you need to target these specifically, although it would be intereting to add library support so that these Conventions are not so necessary. The most interesting of those would be explicit georefrencing, of which coordinate variables / dimension scales are a good start. If theres intereste in adding a more complete coordinate system specification, i would be all for it.